Week 15

Question 1

What are the assumptions of a t test? (4)

Answer 1

• Data is randomly sampled from whole population of interest
• Independent variable is categorical
• Dependent variable is continuous
• Outcome is normally distributed

Question 2

When may it be unsuitable to use a t test? (4)

Answer 2

When assumptions aren’t met, like when data is not normally distributed or sample size is small

When there are more than 2 groups to compare

Question 3

What are the different types of skew? (6)

Answer 3

Positive skew/ right skew:
Low number of high values
Mean becomes higher and more to the right of the median

Negative skew/ left skew:
Low number of low values
Mean becomes less than and more to the left of the median

Question 4

How is skewness measured? (3)

Answer 4

Using Pearson skewness of coefficient
Positive= positive skew
Negative = negative skew

Larger the PSOC, the more skewed: if more than 1 or less than -1 the data is probably not normal

Question 5

What is kurtosis? (3)

Answer 5

Measures tail extremity and represents the presence of outliers

Decrease in kurtosis: decrease of presence of outliers
Increase in kurtosis: increase of presence of outliers

The larger the number, the less normally distributed the data
If Kurtosis is less than -1 or more than 1 then the distribution is not normal

Question 6

When are positive and negative skews common in clinical investigations?

Answer 6

Positive skew:
When a normal amount is low or zero
Or when time is a short duration

Negative skew:
When a normal amount is above a certain value
When time is long but but infinite

Question 7

What can be done if continuous variable is not normally distributed? (3)

Answer 7

Transform the data:

• Use logarithmic
• Use ordinal categories so chi squared test can be used

Use non parametric statistical tests

Question 8

What is the difference between nominal and ordinal variables?

Answer 8

Nominal:
categorical variables that do not have a particular order

Ordinal:
categorical variables that have an order

Question 9

What is the difference in null hypothesis for a parametric and non parametric statistical test?

Answer 9

H0 in parametric statistical test
No difference in mean

H0 in non parametric statistical test
No difference in median